A system for determining optimal treatment of diabetic retinopathy was devised on the basis of fuzzy logic. In a previous study we devised a new retinopathy scoring system from the multivariate analysis of fluorescein angiography findings. Using this scoring system we expressed severities as language terms and assigned them membership functions which showed grade (0-1). Next, the optimal treatment was determined by defuzzification of the membership functions given to treatment programs corresponding to the severity of retinopathy. This decision system is shown as a nomogram. Users of this system can get the grade for choosing optimal treatment after they input the retinopathy score into the nomogram. Assessment of this fuzzy decision system was performed by changes in the extent of dye leakage (Le) and neovascularization (NV) in 86 eyes of 48 diabetic patients followed up for over 12 months. When the grade of choice was over 0.5, the evaluation of choosing optimal treatment was 74% by Le and 68% by NV. Le improved but NV did not improve when the grade of choice was elevated to a higher level.